[
https://issues.apache.org/jira/browse/HIVE-24415?focusedWorklogId=531044&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-531044
]
ASF GitHub Bot logged work on HIVE-24415:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 05/Jan/21 07:15
Start Date: 05/Jan/21 07:15
Worklog Time Spent: 10m
Work Description: mustafaiman commented on a change in pull request #1701:
URL: https://github.com/apache/hive/pull/1701#discussion_r551753955
##########
File path:
ql/src/java/org/apache/hadoop/hive/ql/exec/tez/DynamicPartitionPruner.java
##########
@@ -100,37 +99,19 @@
private int sourceInfoCount = 0;
- private final Object endOfEvents = new Object();
-
private int totalEventCount = 0;
- public DynamicPartitionPruner(InputInitializerContext context, MapWork work,
JobConf jobConf) throws
- SerDeException {
- this.context = context;
- this.work = work;
- this.jobConf = jobConf;
- synchronized (this) {
- initialize();
+ public void prune() throws SerDeException, IOException,
InterruptedException, HiveException {
+ if (sourcesWaitingForEvents.isEmpty()) {
+ return;
}
- }
-
- public void prune()
- throws SerDeException, IOException,
- InterruptedException, HiveException {
- synchronized(sourcesWaitingForEvents) {
Review comment:
Earlier it was neccessary because addEvent and DynamicPartitionPruner
constructor were both modifying sourcesWaitingForEvents. Now, that set is only
modified by DynamicPartitionPruner#prune (and the methods it calls).
DynamicPartitionPruner#prune is called from HiveSplitGenerator's init method
which is called only once by Tez
(https://github.com/apache/tez/blob/849e1d7694cdfd2432d631830940bc95c6f26ead/tez-dag/src/main/java/org/apache/tez/dag/app/dag/RootInputInitializerManager.java#L281).
Concurrent part is addEvent calls. addEvent calls just queue the events and
not touch any mutable state. Therefore further synchronization is not necessary.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 531044)
Time Spent: 40m (was: 0.5h)
> HiveSplitGenerator blocks Tez dispatcher
> ----------------------------------------
>
> Key: HIVE-24415
> URL: https://issues.apache.org/jira/browse/HIVE-24415
> Project: Hive
> Issue Type: Improvement
> Reporter: Mustafa İman
> Assignee: Mustafa İman
> Priority: Major
> Labels: pull-request-available
> Time Spent: 40m
> Remaining Estimate: 0h
>
> HiveSplitGenerator does a lot of heavyweight operations in its constructor.
> These operations block AsyncDispatcher in Tez
> [https://github.com/apache/tez/blob/989d286d09cac7c4e4c5a0e06dd75ea5a6f15478/tez-dag/src/main/java/org/apache/tez/dag/app/dag/RootInputInitializerManager.java#L141]
> . We should move bulk of initialization out of constructor.
> The only reason of setting up everything in constructor is
> DynamicPartitionPruner. We can buffer incoming events in HiveSplitGenerator
> until dynamic partition pruner is initialized.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
